claritystorm/nhtsa-fars-1975-2023
NHTSA FARS Fatal Crashes 1975–2023 One-time dated snapshot. Fatal crash, vehicle and person records in CSV and Parquet. Verified coverage: 1975–2023; all 49 years present in each table. Records: 1,836,052 crashes; 8,344,160 rows across three tables. This repository contains a 1,000-row public sample, not the full package. A sample does not establish complete historical coverage. Limitations Crash, vehicle and person rows represent different units; do not add them… See the full description on the dataset page: https://huggingface.co/datasets/claritystorm/nhtsa-fars-1975-2023.
NHTSA FARS Fatal Crashes 1975–2023
One-time dated snapshot.
Fatal crash, vehicle and person records in CSV and Parquet.
Verified coverage: 1975–2023; all 49 years present in each table.
Records: 1,836,052 crashes; 8,344,160 rows across three tables.
This repository contains a 1,000-row public sample, not the full package. A sample does not establish complete historical coverage.
Limitations
- Crash, vehicle and person rows represent different units; do not add them to estimate crashes.
- 74,669 crash rows contain non-null coordinates outside valid latitude/longitude bounds. Filter invalid values and documented source sentinels before mapping.
- Coordinates are missing for many older records. No guarantee of complete geospatial coverage.
- Join tables on year and stcase; include vehno for vehicle joins. The 6,059 person rows without a matching vehicle are all coded as occupants of vehicles not in transport; a parked/working-vehicle table is not included.
- The current official 2023 files have 37,769 crash rows versus 37,654 here. This snapshot has not been reconciled to subsequent source revisions.
Package upload: 2026-03-30. Listing reviewed: 2026-09-19. These dates do not extend source coverage. This review did not refresh the data.
Existing package tables
Counts are per table, not independent events. Full field names and types are on the dataset page.
Inspect the public CSV
import pandas as pd
df = pd.read_csv("sample_1000.csv")
print(df.shape)
print(df.columns.tolist())
print(df.head())Availability
The dated snapshot is available for a one-time $99 purchase, subject to its documented limitations and license terms. Future updates are not included.
New annual update subscriptions and bundle purchases are paused. Existing customers retain their purchase download access and can contact support@claritystorm.com for assistance.